PDA

View Full Version : overlap integrals for empirical data



choschech
04-13-2010, 11:27 AM
Hi,

I have two sets of data each originating from a different probability density function for a continuous stochastic variable (I do not know how these functions look like exactly). I am interested in the overlap integral of the two probability densities. I suppose the simplest way would be to bin the data and then consider probabilities for these bins. I wonder, however, whether there is a more sophisticated method... maybe something like considering overlap integrals of Gaussians that are centered on the data points?


Best

choschech

AtlasFrysmith
04-13-2010, 11:44 AM
That sounds a lot like kernel density estimation to me: http://en.wikipedia.org/wiki/Kernel_density_estimation

choschech
04-13-2010, 08:13 PM
Thanks a lot AtlasFrysmith! That sounds exactly like what I need!